What if it already went to a client?

Tell them today, with the corrected version attached. Discovery by the client is a different-severity event.

Will I lose access to AI tools?

Less likely if you bring the fix with the disclosure.

What if it was in a legal or medical context?

Escalate to legal or compliance immediately and stop drafting the response yourself.

← All playbooks

Tier 4: Total Clusterfuck

What do you do if AI-generated content you shipped contained made-up facts?

The short answer

Pull or correct the artifact immediately and identify every place it was distributed. Tell your manager and the recipient that the content contained unverified claims, name which ones, and state the verification step you're adding. Do not lead with 'the AI did it' — the fabrication is a process failure and the process was yours.

What's actually at stake

01Fabricated citations spread further than the correction does; distribution mapping matters more than the apology.
02Blaming the model reads as an admission that you don't review your own output.
03A named verification step is what makes people comfortable letting you use the tool again.

Do

· List every channel the content reached.
· Name the specific false claims, not 'some inaccuracies'.
· Commit to a concrete verification step.

Don't

· Do not lead with the tool's name.
· Do not quietly swap the file and say nothing.
· Do not promise to stop using AI — nobody believes it and it's the wrong fix.

A draft you can send now

Heads up: the [doc/post] I sent contained two claims I can't source — specifically [A] and [B]. I've corrected the version at [link] and notified [recipients]. Going forward I'm citing-and-checking every factual claim before anything leaves my hands.

Tailor this to my exact disaster →

Questions people actually ask

What if it already went to a client?: Tell them today, with the corrected version attached. Discovery by the client is a different-severity event.
Will I lose access to AI tools?: Less likely if you bring the fix with the disclosure.
What if it was in a legal or medical context?: Escalate to legal or compliance immediately and stop drafting the response yourself.

Take it further

Ask Cris about this →

Talk it through with the crisis bot using your actual details.

Grade what you already sent →

Incident Autopsy scores the damage and rewrites it.

Rehearse the conversation →

Spar with a skeptical manager, exec, or client before the real one.

Tier 4: Total Clusterfuck explained →

Money, customers, or a very senior person's calendar is involved.

Related incidents

All CYABot tools →